Various methods and apparatus are under current research and development for transferring, using a charged particle beam, a pattern defined by a reticle or mask onto a sensitized substrate by microlithography. Representative charged particle beams used in such systems include electron beams and ion beams. Electron-beam systems have been the subject of most such effort; hence, the following summary is in the context of electron-beam systems.
Charged-particle-beam (CPB) microlithography systems, such as electron-beam writing systems, offer tantalizing prospects of improved accuracy and resolution of pattern transfer, but exhibit disappointingly low throughput. Consequently, much contemporary research and development has focused on overcoming this disadvantage. Examples of various conventional approaches include "cell-projection," "character projection," and "block projection" (collectively termed "partial-block" pattern transfer).
Partial-block pattern transfer is used especially whenever the pattern to be transferred to the substrate comprises a region in which a basic pattern unit is repeated many times. For example, partial-block pattern transfer is used generally for patterns having large memory circuits, such as DRAMs. In such patterns, the basic pattern unit is very small, having measurements on the substrate of, for example, (10 .mu.m).sup.2 (i.e., 10 .mu.m.times.10 .mu.m). The basic pattern unit is defined on one or several exposure units on the reticle and the exposure units are repeatedly exposed many times onto the substrate to form the pattern on the substrate. Unfortunately, partial-block pattern transfer tends to be employed only for repeated portions of the pattern. Portions of the pattern that are not repeated are transferred onto the substrate using a different method, such as the variable-shaped-beam method. Therefore, partial-block pattern-transfer has a throughput that is too low, especially for large-scale production of integrated circuits.
A conventional approach that has been investigated in an effort to achieve a higher throughput than partial-block pattern-transfer methods is a projection microlithography method in which the entire reticle pattern for a complete die (or even multiple dies) is projection-exposed onto the substrate in a single "shot." In such a method, the reticle defines a complete pattern, such as for a particular layer in an entire integrated circuit. The image of the reticle pattern as formed on the substrate is "demagnified" by which is meant that the image is smaller than the pattern on the reticle by a "demagnification factor" (typically 4:1 or 5:1). To form the image on the substrate, a projection lens is situated between the reticle and the substrate. Whereas this approach offers prospects of excellent throughput, it to date has exhibited excessive aberrations and the like, especially of peripheral regions of the projected pattern. In addition, it is extremely difficult to manufacture a reticle defining an entire pattern that can be exposed in one shot.
Yet another approach that is receiving much current attention is the "divided" or "partitioned" projection-exposure approach that utilizes a "divided," "partitioned," or "segmented" reticle. On such a reticle, the overall reticle pattern is subdivided into portions termed herein "exposure units." The exposure units can be of any of various types termed "subfields," "stripes," etc., as known in the art. Each exposure unit is exposed individually and sequentially in a separate "shot" or scan. The image of each exposure unit is projection-exposed (typically at a demagnification ratio such as 4:1 or 5:1) using a projection-optical system situated between the reticle and the substrate. Even though the projection-optical system typically has a large optical field, distortions, focal-point errors and other aberrations, and other faults in projected images of the exposure units are generally well controlled. Although divided projection-exposure systems provide lower throughput than systems that expose the entire reticle in one shot, divided projection-exposure systems exhibit better exposure accuracy and image resolution.
In divided projection exposure, it is necessary to achieve very accurate alignment of the reticle with the substrate to ensure that the images of the exposure units are positioned at the respective locations on the reticle with extremely high accuracy. To such end, an operation termed "mark detection" is performed such as during calibration of the optical system and when aligning the substrate with the reticle before exposing an exposure unit onto the substrate. During mark detection, an image of one or more "upstream" marks provided on the reticle or other location on the reticle stage is projected onto a corresponding "downstream" mark provided on the substrate or other location on the substrate stage. The marks are scanned relative to each other to determine the relative positions of the marks.
Systems designed for high-resolution pattern transfer, such as the divided projection-exposure system summarized above, employ very large acceleration voltages such as between the CPB source and the reticle. To achieve the requisite high accuracy of mark detection, either mark scanning must be performed relatively slowly or a large number of scans must be performed. Consequently, the cumulative beam energy that strikes the marks and their immediate surrounding area is very high. This energy usually is dissipated as localized heating which elevates the temperature and causes thermal deformation of the vicinity of the marks. Such deformation degrades the accuracy with which mark positions can be determined, reduces calibration and alignment accuracy, and reduces the accuracy with which images of exposure units on the substrate can be stitched together. The resulting devices manufactured under such conditions exhibit a higher incidence of defects such as shorts, opens, and non-uniform resistance values.